ARM4SNS

• ARM4SNS Web Ressources

Introduction

Connected to the internet you have a lot of opportunities to interact with strangers. Some of the interactions are funny or informative and some of them are profitable and all involve risks. A reputation system gives people information about others's past performance. It helps people decide who to trust, encourages people to be more trustworthy and discourages those who are not trustworthy from participating.

• Resnick, Paul, Zeckhauser, Richard, Friedman, Eric, and Kuwabara, Ko. Reputation Systems. Communications of the ACM, 43(12), December 20000, pages 45-48 (it gives a nice introduction what are reputation systems good for)

In general, what is Reputation Information good for?

• Value Preposition: To WHOM does reputation provide WHAT value?
  • Q: Where can the availability of reputation information help to make things better?
  • A: Generally, reputation helps to decide wheter or not to interoperate with a unknown peer. More specifically,
    • it helps to maximize the return on an fixed investment of resources (time, monney, attention) (see Example)
    • it helps to control access to uncritical (e.g. low cost) resources (see Example)
    • it enables relaxed policies in certain business processes (see Example)
  • Q: How does it help there (what does 'better' mean)?
• Limitations: What can Reputation Information NOT provide?
  • Access control for critical resources
  • Reputation is a statistical guess for future behaviour on the basis of the past. Therfore only statistical estimates for the behaviour of a sufficient large ensemble of peers can be taken. For one individual a strict prognosis is impossible.

What is Reputation?

• a 1,2,n dimensional value?
• a discrete or a rational number?
• comparable, combinable?

Userfriendly submission/use of Reputation Information

• Tools for making submission of reputation information user friendly
• Tools for making use (application) of reputation information user friendly
• Acceptability of Reputation Information by end users

Anonymously Supplying/Storing/Using Reputation Information

• using a tusted reputation provider who must keep all information secret
• make it algorithmically impossible to derive a reputation information supplier's / user's identiy.

Toolset ... for building Reputation-based Solutions

Architecture for a Reputation Information Management System

• building blocks for building a reputation management system
  • what functions do the building blocks provide?
  • how do they depend on each other?
  • alternative building blocks for similar functionality
• sample scenario for illustrating building block's functionality and interactions

Method for integrating Reputation Information into a product/solution

• Given an application and given a reputation management system, are there common techniques for integrating the two?

Working Packages

• which building block?
• which requirements?
• expected outcome?

State of the art

• Literature survey
  • The Beta Reputation System (2002), Audun Jøsang, Roslan Ismail, In Proceedings of the 15th Bled Electronic Commerce Conference, Bled, Slovenia, June 2002 (PDF)
    • Posteriori probabilities of binary events can be represented as beta distributions. Combining feedback resulting from an e-commerce transaction is not the same as statistical observation of a binary event, because an agent's perceived satisfaction after a transaction is not binary. Instead positive and negative feedback is given as a pair (r,s) of continuous values where r reflects the degree of satisfaction and s reflects the degree of dissatisfaction. Discounting (highly reputed agents carry more weight when they give feedback) and forgetting (old feedback may not be relevant) of the reputation values are also used.
  • Trustdavis (2005), DeFigueiredo, Barr, (PDF)
    • Newcomers have to pay a deposit for references. The deposit is used to save transactions. Pseudonyms are more expensive.
  • XENO Trust (2003), Dragovic, Kotsovinos et al., (PDF)
    • Architecture for the storage, retrieval and aggregation of reputation information. Used in the XenoServer Open Platform: a public infrastructure for wide-area computing.
  • EigenTrust (2003), Kamvar, Schlosser et al., (PDF)
  • SPORAS/HISTOS (1999), Zacharia, Maas, (PDF)
    • SPORAS is a simple reputation mechanism for losely connected online communities and an evolved version of eBays or Amazons reputation mechanisms. It introduces the notion of the reliability of the reputation. It also has a forget factor to consider the most recent impressions. HISTOS is for highly connected online communities and a more personalized reputation system, where reputation depends on who makes the query and how that person rated other users. User anonymity through pseudonyms are allowed in both systems.
  • DCRC/CORC (2003), Gupta, Judge et al., (found at Citeseer)
  • Regret (2001), Sabater, Sierra, (PDF)
    • They claim that reputation has different dimensions (individual, social, ontological). Regret gives more relevance to recent impressions and computes also the reliabilty of a reputation.
  • TrustNet (2000), Schillo, Funk et al. (found at Citeseer)
  • TrustMe, Singh, Liu, (found at Citeseer)
    • Secure and anonymous protocol for Trustmanagement. Uses public key cryptography mechanisms. Each peer has a couple of public-private key pairs. Trust value of a peer (A) is randomly assigned to another peer (Trust holding agent=THA) at bootstrapping. The trust holding responsibilities are equally distributed amongst the peers and are unknown to all peers including peer (A). A peer (B) broadcasts a trust query for peer (A), the THA replies with the trust value.
  • A Survey of Trust and Reputation Systems for Online Service Provision, (2005), Jøsang et al. (PDF)
    • They provide some details for eBay, AllExperts, Advogato, Epinions, BizRate, Amazon, Slashdot, Kuro5in, PageRank and they describe various principles for computing reputation and trust measures.
• Conferences to watch
• Companies active in this area
  • Open Ratings http://www.openratings.com/
• Examples (to move in a restricted area?)
  • PageRank taken from "A Survey of Trust and Reputation Systems for Online Service Provision", (2005), Jøsang et al. (PDF) (see PageRank)
    
    • $P$: set of hyperlinked webpages
      
    • $u,v$: webpages in P
      
    • $N^{-}(u)$: set of webpages pointing to u
      
    • $N^{+}(v)$: set of webpages that v points to
      
    • the PageRank is: $R(u)=cE(u)+c\sum _{v\in N^{-}(u)}{R(v) \over {|N^{+}(v)|}}$ (1.)
    • $c$ is chosen such that $\sum _{u\in P}R(u)=1$
      
    • $E$ is a vector over $P$ corresponding to a source of rank and is chosen such that $\sum _{u\in P}E(u)=0.15$
      
    • first term of function (1.) $cE(u)$ gives rank value based on initial rank
      
    • second term of (1.) $c\sum _{v\in N^{-}(u)}{R(v) \over {|N^{+}(v)|}}$ gives rank value as a function of hyperlinks pointing at $u$